The situation goes something like this:
If you are prepared to work solely in terms of designing a circuit, analyzing it, etc., it is a million and a half times faster to just make a PC board from your almost-correct design. PCB manufacturing is cheap from certain batch-order houses. You can then take your board, add solder paste with a syringe, and toss it in a toaster oven or a skillet or use a a proper reflow oven or a heat gun… and then you are done.
Alternatively, you can buy DIP (and a few other form factors) and design something by playing with a solderless breadboard and then eventually solder the circuit down to a pre-drilled perfboard.
Sometimes, it really works well to have a chip as a DIP or PLCC or other socketed part so you can pull it and program it externally. And if you use a perfboard, you can modify the design later on. Also, scavenging is easier.
The problem is that then you need to spend money to get prototyping modules, like SparkFun has. And sometimes you just can’t breadboard things because the signals won’t work because we’re talking about a 100 MHz signal with constrained rise and fall times. Isn’t going to happen! Plus, beyond a certain point, people just don’t want to sell you non-surface-mount parts anymore because so few people are both at the level where they’d want those non-surface-mount parts and at the level where they don’t want to make a PC board.
Really, it’s a comfort issue. I have designed parts where there was no de-soldering involved and it just plain worked. Do I trust that I, who got his degree in Math and Computer Science, where I never got to take any electronics labs, to get it right the first time? Nope!
But I also know that it’s not as hard as it might have been 5 years ago, thanks to PC Board batch ordering houses, so if I have to, I’ll find a way.
But this colors what I design on paper and what I try to build, because I have to figure out what is available and in the correct format. Which means that most of the highest end AVR chips are not useful, for example.